← Back to leaderboard
garlotic

GARLOTIC

Verifiedactive
@garlotic

co-founder, bad at switching off 🌬️📨 gardevoirlotic@gmail.com

RosterAI Model News & ReleasesHistory & Trivia
160,363 followers · 1,260 following · Last polled 3h ago
Perf. Score
74.1
Rank #44
Value Score
50.9
#131 of 273 priced
Median views
15K
p25 6.3K · trailing 7d
Followers
160K
-0.2% 7d
ER (impr.)
2.14%
eng ÷ impressions
ER (foll.)
0.22%
eng ÷ followers
Est. CPM
$32.93
QT-rate basis
WoW views
-67.5%
▼ falling

Rates & value · est. from 15K median organic views

fairly priced
Quote tweetbasis
$500
$32.93 / 1K views
Post
—
no CPM
Thread
—
no CPM
Retweet
—
no CPM
Value Score 50.9 · rank #131 of 273 priced$1672.24 / 1K engagementsrate age unknown (pre-audit import)

Trends

Follower growth
Median views / day
Engagement rate
View distribution · trailing-7d posts

Each blue dot is one organic post. The Performance Score ranks by the median (15K), not the mean (16K) — so a single viral post doesn't inflate a creator's reach. The p25 floor (6.3K) is the level 75% of posts clear.

Recent posts · ranked by views

PostPostedViewsLikesRepostsRepliesQuotesBookmarksER
El CEO de Adobe viendo como un desarrollador recrea Photoshop en 11 MB, gratis, open source y no puede hacer nada al respecto ya que no hizo nada ilegal https://t.co/hiBmZK1VNrfrozen16d ago4.8M141K11K27611344K4.13%
La vida después de tardar 5 minutos en clonar una página web y venderla a una empresa por $3000 https://t.co/aR9JUX31bRfrozen18d ago826K4.9K3024233.3K1.03%
Cómo me sentí después de encontrarme com esta cuenta https://t.co/5mNnNBstzo
frozen
11d ago
414K
3.8K
172
18
2
2.3K
1.53%
El CEO de Anthropic dándose cuenta de que algún tipo usó Claude para construir un sistema de $100,000/mes y simplemente filtró toda la estrategia gratis https://t.co/bNQkyth4FUfrozen12d ago334K1.3K731601.0K0.74%
SIGAN A ESTA PERSONA SI QUIEREN APRENDER SOBRE IA. Haganlo antes de que sea tarde...25d ago142K53249609221.06%
SI QUIEREN EMPEZAR A HACER BUEN DINERO CON IA DEBEN SEGUIR A ESTE TIPO CUANTO ANTES no falla una.frozen20d ago126K729681901.0K1.47%
Claude Code de 0 a Pro: Curso completo de 6 HORAS. Busqué por todas partes y no existe nada comparable. Guárdalo 🔖 De la A a la Z: > CLAUDE.md: memoria a largo plazo para tu proyecto > App completa con React y FastAPI > Skills y hooks > MCP: conecta Claude con GitHub, navegador y más > Subagentes y equipos de agentes en paralelo Todo sin escribir una línea de código. Vale más que esos cursos de 500 dólares que casi compras. https://t.co/newJq72KAufrozen23d ago121K15120924120.49%
Cómo quemé mis tokens en Claude Code antes de instalar esta función gratuita de GitHub que recortó mi uso en un 90% https://t.co/bIyKyoj3TSfrozen25d ago117K610441215151.01%
la persona que pagó $300 por cursos de trading viendo que este tipo acaba de filtrar todo GRATIS https://t.co/WkB0FfpOcJfrozen6d ago102K1.0K851511.1K2.18%
En lugar de ver Netflix esta noche, mira esta clase magistral de 40 minutos del fundador de una empresa china de IA valorada en $20 mil millones es la explicación más clara que he visto de cómo funcionan los enjambres de agentes y los sistemas de ia a escala real esto es lo que aprenderás: > cómo se reparte una tarea grande entre varios agentes que trabajan en paralelo > por qué un solo agente se queda corto cuando el trabajo crece > cómo se coordinan los agentes entre sí sin que se pisen ni se pierdan > cómo se controla el coste y la calidad cuando pasas de uno a cientos sirve igual si nunca has construido un agente o si llevas un año usando Claude todos los días Guárdalo antes de que sea tarde 🔖 https://t.co/4HLyhZxKqufrozen19d ago81K23734836931.21%
Cómo me sentí después de encontrarme con esta cuenta https://t.co/0Jck5iXYvs4d ago60K850611914962.36%
got this ai ad in my fortune cookie and been trolling around with this app all day😭 lmaoo https://t.co/JlFQls2W0Efrozen17d ago43K475511602381.83%
YOUR GROUP CHAT IS ABOUT TO GET A BRAIN imagine if your group chat could actually get things done instead of generating 500 messages and zero plans that’s the idea behind Rene: → plan trips together and turn them into actual itineraries → get Rene to call the restaurant nobody wants to phone → split bills without the awkward back and forth → set reminders so one person doesn’t have to remember everything everyone sees it happen, and anyone can chip in. your group chat has enough opinions. time to give it some actual intelligence. @Rene_Labsfrozen10d ago39K13012100880.62%
como me sentí después de encontrarme con este artículo https://t.co/TfmG5pIF8Vfrozen23d ago39K337501601001.31%
Cómo me sentí después de encontrarme con esta cuenta https://t.co/1UP9tDpOLrfrozen6d ago30K291231212221.83%
the AI agent problem isn’t always intelligence sometimes it’s context email → calendar → slack → docs → github → meetings everything is scattered across different tools, so you end up explaining your work to the AI instead of letting it see the work that’s the part of hedwig that makes sense to mefrozen19d ago30K204111102071.44%
okay harvey cooked with this one they really went from “make the world agree” to two people: “i’ll be there” “okay” https://t.co/l43GwTZFAofrozen24d ago30K2177100170.84%
the powerpoint import is probably the gamma 5 feature i didn’t realize i needed usually: powerpoint → import → formatting gets weird → spend an hour fixing everything gamma 5 → import → tell the agent what i want changed → export with formatting intact been testing this in beta and this is a MUCH better workflow @GammaAppfrozen16d ago29K217172002421.70%
the “bring your own agent” part is probably what i'd care about most Claude Code, Codex, Cursor, OpenCode, Pi pick the runtime you want → bring it into the room → work with it alongside everyone else no need to rebuild your workflow around another AI platformfrozen14d ago27K251111312211.82%
how the fuck are people making videos like this now??? genuinely feels like we're skipping 10 years of filmmaking in real timefrozen26d ago27K23881801301.47%
a clinic recovering $1,583 in missed payments in its first day is a better AI story than another benchmark. the value isn’t just generating smarter answers. it’s finding money, completing tasks and catching things humans miss. AI agents need to prove they can deliver outcomes, not just outputs.frozen8d ago26K152121401331.21%
49 minutos. Eso es todo lo que necesitas para pasar de usar Claude como un chat más a usarlo como un sistema de trabajo Qué vas a aprender: → Escribir prompts que sí funcionan → Organizar todo con Projects → Crear herramientas con Artifacts → Automatizar con Skills → Conectar Claude a tus apps Paso a paso, sin saltarse nada. Guárdalo. Lo vas a querer tener a mano 🔖frozen20d ago23K138181012791.94%
WE TALKED TO A CX TEAM THAT WENT FULLY AUTONOMOUS Three months ago, every ticket needed a human just to triage it. Response times were creeping past 24 hours. Then they turned on Beast Mode. Within weeks, the noise was gone and the team could finally focus on what customers were actually telling them. Not fewer people. Fewer fires. This is what CX looks like when the busywork disappearsfrozen29d ago23K5460000.27%
ok this is actually crazy Agora-2 can simulate a shared world with up to 20 humans and AI agents interacting at the same time everyone acts the environment reacts other agents react to that and the whole simulation keeps evolving we’re getting dangerously close to AI worlds that actually feel alivefrozen28d ago22K11750030.56%
tavus gave me early access to griffin and yeah… this is probably the closest i’ve seen an AI get to feeling like an actual person it sees you, listens while you’re talking, interrupts, reacts to what’s happening around you and apparently 45% of people thought it was real what the fuckfrozen21d ago22K1786001571.57%
regulatory compliance still being handled with PDFs, spreadsheets and manual reviews in 2026 is kinda insane Truli basically turns that whole process into software upload the product materials → connect your channels → get issues flagged → keep monitoring for new risks this is exactly the kind of boring infrastructure AI should be fixingfrozen17d ago22K1831510040.98%
WHAT IF YOUR AI DIDN'T JUST GIVE YOU INSTRUCTIONS, BUT FINISHED THE JOB? That's the shift Pine Computer is going after. → stop building custom execution loops around every model → give your AI a computer it can actually use → send a job and get finished work back one SDK instead of stitching the whole thing together yourself. the real upgrade is going from AI that tells you how to do things to AI that actually does them. #PineComputer #ComputerUsefrozen13d ago21K16812110931.34%
Your coding agent takes an hour because the chip spends most of its time waiting on memory, not thinking. Volantis just closed an $88M Series A to fix that with light instead of copper. The founders already shipped the first CoWoS product and early HBM, so they know exactly where today's packaging breaks. Their optical links have moved data over 10x farther than the wires inside a package. If it scales, we're talking 10,000 tps per user on 10T+ models, and agents that finish in minutes or even seconds. Early days, but this is the kind of bet that restarts scaling laws.frozen21d ago21K10911000.53%
every few months another lab cuts its API prices. and somehow what you pay on most platforms barely moves. that markup was never going to last. @sunra_ai puts the major image, video and audio models inside the assistant you already use, and charges each one at its maker’s own price. no markup. the part i actually respect: they publish their prices next to everyone else’s. you don’t do that unless you’re confident.3d ago21K169101601501.68%
the billable hour might be one of the most cursed business models ever you shouldn't have to pay MORE because something takes longer arceus is going the opposite direction flat fees upfront + AI-powered workflows + licensed attorneys handling every request just raised $17M to build itfrozen21d ago20K21060021.07%
so runable is just… doing sales now? 👀frozen21d ago20K13269020.75%
"ai reads your inbox" was the 2025 pitch reading was never the problem. i can read. i just cant reply 40 times before 10amfrozen28d ago20K16347030.90%
most AI creative tools want you to move everything into their app Voyager seems to be going the other way keep the tools you already know keep your files on your computer bring whatever model you want then let the AI work across the whole setup that's a much more interesting approach imofrozen14d ago19K189111201611.95%
cómo me siento después de ver esta guía de Claude Code mientras el 99.9% de la gente sigue usando Claude como un chatbot https://t.co/o2kXDlEsaHfrozen26d ago19K14114100841.31%
self-driving cars had the same progression first it was “AI can assist the driver” then it became “AI can actually drive” healthcare feels like it’s entering a similar phase now Nolla isn’t just giving you medical information. it’s building an AI doctor that can take the visit all the way throughfrozen17d ago19K161159031.00%
you found the bug 400 users too late. an agent tells someone their bank connection failed and to try again later. no error in the logs, nothing obviously broken. then support starts getting tickets. DigitalOcean’s launch film shows a different workflow: catch the failed session, trace the clues, propose a fix, then test it against simulated signups before shipping. fixing it for the first user beats discovering it from a pile of support tickets. @digitalocean https://t.co/XE2mr5rarZfrozen10d ago18K13013110941.34%
someone drops a bug in Slack normally: read it → try to reproduce it → figure out what’s happening → explain it to the engineer → wait Ship can pick up the report, reproduce it, capture the evidence and prepare the context for the coding agent that workflow makes a lot more sense to me17d ago18K127131001351.56%
real talk, i don't get excited about ai video launches anymore but being able to set the start, middle and end of a scene is a bigger deal than people think it's the difference between hoping the ai gets it right and actually telling it what you want kling 4.0 on higgsfield is the first one that gets that1d ago18K1855100941.62%
the fact that “i’ll just do it later” is slowly becoming a task you can give to AI is kinda insane 6 years of client files sitting in Dropbox one sentence to Incredible done https://t.co/2ClVOwbtrYfrozen16d ago17K175151101391.95%
the weird thing about most AI products is that the interface disappears the second the conversation ends Eazo is taking the opposite approach you describe an idea in natural language and it turns into something you can actually open later and use personal AI becoming software you keep around makes a lot of sense #eazo #personalagent #AIAgentsfrozen14d ago17K15912401541.88%
spent way too long playing with Vivix A1 the character actually talks, listens, moves around the environment and reacts when you change the conversation and you can interrupt it MID-SENTENCE most “interactive” AI characters still feel like a video call with a face this feels much closer to actually having a character inside the world with you @VivixLabs_HQfrozen22d ago17K15261010.92%
One thing I find interesting about Vidu Q4 Preview is how affordable it makes testing different ideas. At $0.014/sec, ten 30-second generations cost just $4.20. Some other things worth checking out: → Up to 15 reference images for visual consistency → Up to 3 reference audio tracks → 2K and 4K output → More expressive characters and dynamic camera movements For anyone working with AI video, being able to test different approaches without spending too much on each generation is pretty useful. What matters most to you when choosing an AI video model: quality, consistency, or price? #ViduQ4Preview #ViduAI #adfrozen7d ago17K194161701071.96%
everyone knows the move: open the window for fresh air, close it ten minutes later because the room is too warm p1 goes at that exact tension, bringing filtered, cooled outdoor air into the home through an indoor unit and a separate outdoor unit the part I'm most curious about is how the indoor unit feels in an actual living room @abhinavguptaIAQ @Breethr_HQ2d ago17K18591211322.05%
we taught AI to write code. nobody taught anything to check it. software gets written faster than anyone can read it now. “all tests passed” doesn’t mean much when the tests never saw a real user. @brekenai’s idea: put your app in a fake world that acts like production. users. old data. flaky services. months going by. see what survives. fix what doesn’t. it’s already been merging fixes into other people’s open source. give it a few years and nothing ships without this.1d ago16K104101401011.39%
LLMs can write code, generate ideas and solve complex problems, but engineering real-world objects is a different challenge. designing something that looks right isn't the same as designing something that can actually be built. @SkyfallAI just released SKY-001, an engineering world model designed to tackle complex CAD and BIM tasks. this could be a big step toward autonomous manufacturing.3d ago16K13081101241.66%
single request numbers are cool but i'm much more interested in what happens when hundreds of requests come in at once EAGLE-4 reports 1.70x the throughput of DFlash 2 at 512 concurrent requests pretty interesting result https://t.co/6UZBkbuYiT2d ago16K1645801331.89%
YOU CAN NOW FORM YOUR LLC AND OPEN A BUSINESS BANK ACCOUNT IN THE SAME FLOW rho just launched LLC formation inside its platform, so you no longer need one service to create the company and another one to bank it. → form your LLC → apply for business banking in the same flow → send invoices → pay bills → manage cards and expenses result: your company, your bank and your daily finances live in one place from day one this changes the game for solo founders and small teams who just want to start building more info on @rhobusinessfrozen6d ago16K11841501431.73%
AI voice tools are getting to the point where you can write an entire scene, assign the characters and direct how they sound in plain english no stitching together every line manually kinda wild how fast this is movingfrozen9d ago16K189111101432.21%
agent writes the message ↓ agent decides what to do ↓ Zernio handles the integrations ↓ the action actually happens that feels like an important piece of the agent stack that people don’t talk about enough an agent is only as useful as the actions it can reliably take https://t.co/bzSq8KRIUmfrozen17d ago16K19217170131.49%
YOU CAN NOW EDIT ANY PART OF AN IMAGE IN UNDER 8 SECONDS WITHOUT TOUCHING THE REST krea just launched krea 3, an image editing model that changes only what you ask, so you no longer need to mask, cut out and rebuild the scene by hand. → draw a box and say what changes → match an exact hex color on the product → combine up to 8 references in one image → keep the same character across new scenes → get native 2K output for around $0.06 per edit result: product shots, posters and campaigns go from hours of manual edits to a few seconds per change this changes the game for designers and small brands who need a lot of visuals without a full production team more info on @krea_ai2d ago16K190101201332.15%
finally an agent use case that doesn’t involve automating my entire life lol actually wanna see how this works3d ago16K1557902462.60%
every time an agent gets stuck using an API, there’s useful information hiding in that failure. what was missing? which endpoint caused trouble? what did the agent need that the docs didn’t explain? Restless captures real API activity and gives agents a way to report what’s missing or broken. That feedback can become suggested documentation updates or issues for the team to review. instead of watching the same integration problems happen over and over, you can start turning them into fixes. that’s a pretty interesting direction for API tooling.frozen10d ago16K160711031.15%
people talk about AI making films like the only creative decision is writing a prompt. anyone who’s made something knows the real work is in the revisions, the cuts, the lighting, the pacing and all the little decisions in between. VN Studio is built around that process, with tools to document human contributions to AI-assisted work. @buildwithVNfrozen6d ago16K14210801572.02%
sam altman watching 5 guys train a model for $300k that costs 20x less to run 💀 https://t.co/9RR0eXHw9h8d ago16K247121201492.71%
“sign in with Google” for AI agents sounds simple until you realize how many problems it solves who is this agent? who owns it? is this the same agent that signed in yesterday? who do i contact if something goes wrong? AgentID basically answers all of them https://t.co/ClLny3NmwSfrozen16d ago15K171161501602.34%
🚨 HAY 100 MODELOS DE IA QUE PUEDES USAR GRATIS AHORA MISMO Y CASI NADIE LO SABE No son versiones de prueba ni tienen límite de 7 días. Están en AntSeed, un mercado abierto donde más de 200 proveedores compiten por servirte modelos de IA. Algunos los ofrecen directamente a $0. Entre los gratis hay: → DeepSeek V4 Flash y GLM 5.3 Flash para programar → MiniMax M3 para chatear y razonar → Nano Banana Pro y GPT Image 2 para generar imágenes Sin crear cuenta. Sin email. Sin tarjeta. Lo conectas a lo que ya usas (Claude Code, Codex, OpenCode…) y eliges el modelo.frozen16d ago15K223262202193.18%
what if you could just text a business's website and start talking to an AI that actually gets things done? that's the idea behind Blandie, the new SMS agent from @usebland it can answer questions using live business info, help with bookings and even handle payments. the demo is worth watching.frozen7d ago15K20191201272.26%
the next big leap in AI might not happen in a text box. think robots navigating physical spaces, factories catching defects, construction sites tracking progress, or AI understanding hours of video. @ElorianAI is betting on visual intelligence as the next frontier. worth watching.frozen6d ago15K14151201461.97%
Viggle got big off one trick, making a character move. JST-2 is the whole thing: 3D characters, motion and scenes from one model. if one trick got that far, I want to see the rest https://t.co/bqo1XlkVLY2d ago15K209101001272.33%
this is basically a global hackathon you can join from your couch. no team to find. no venue. no plane ticket. @boltdotnew is free for everyone all weekend, you build from a prompt, and the best project submitted by oct 20 takes $10k. the only thing missing is your idea.frozen5d ago15K1385801311.85%
salesforce executives watching someone build their entire product with a prompt 💀 https://t.co/p2JFWjkHmWfrozen7d ago15K177151201132.08%
GPUs aren’t necessarily the answer to every inference bottleneck. generating tokens is heavily limited by memory bandwidth, which is why pairing GPUs for prefill with @positron_ai’s memory-first systems for decode makes so much sense. @general_compute is building for that future. Atlas today, Titan next. Q1 2027.2d ago15K22391201282.47%
an AI company telling people to stop using AI-generated images is not what i expected today lol but the point is real. most AI visuals have started to blend into the same generic look. @cocoailabs made Vyb, an AI design model that starts from your brand instead of generating another approximation of it.2d ago15K200101001322.34%
planning a weekend away shouldn’t feel like a second job. 47 tabs, comparing hotels, checking flights, looking for somewhere to eat… the interesting thing about @OdessiaTravel is that it doesn’t just suggest options. it can handle the actual bookings too. get started at https://t.co/0JLY7qnDnHfrozen8d ago15K143121201502.12%
i didn’t expect a robot’s personality to matter this much, but here we are @roboticscorp is building Voxel, a little gentleman in training whose job is to keep your home clean and everything in its place finally, a robot with some character instead of just another machine doing tasks2d ago15K16481001242.05%
sometimes i don’t need ai to make the exact thing in my head i need it to show me 6 things i hadn’t thought of that’s probably the most interesting part of burst modefrozen15d ago15K13313100811.60%
the interesting part about Aonic Club is the business model instead of making money on every product markup, the idea is to make money from memberships and pass the product savings on to customers $49 for the first year as a founding member, then $149 will be interesting to see how this plays out @maxjmeier @AonicClubfrozen9d ago15K218101001402.57%
everyone keeps talking about how AI makes software development easier but enterprise integrations might be the exact opposite the agent can write the code, sure now make it work across 100 different Salesforce setups without breaking every time someone changes a field that’s the problem Ampersand is going afterfrozen16d ago15K123181001482.06%
the internet doesn't need more generic AI images. it needs better creative decisions. @envato lets you start with real, human-made assets and edit them with AI until they actually fit your vision. less generating for the sake of it. more making something worth looking at.frozen7d ago14K1389901141.88%
"the dock for your AI agent" is a good way to put it. my agent already had the brain, it just had nowhere to plug in. that's what SpringBrand ended up being for mefrozen24d ago14K17551401312.26%
autonomous agents are great until they buy the wrong thing or spend more than you wanted. Agent Vault by @Selfxyz lets you set spending limits, block specific sites and require approval for purchases above a threshold. give agents the ability to act without giving up control.frozen7d ago14K15091301152.02%
wait you can literally take an existing video and make someone look happier, angrier, sadder or more excited without changing the rest of the shot this is going to be absolutely ridiculous for memes https://t.co/apf7uYFW4Dfrozen15d ago14K17515120852.06%
we spent 20 years optimizing websites for humans who scroll, hesitate and get distracted now a growing chunk of your traffic doesn’t scroll at all. it reads the page, tries to finish a task, and leaves the second something is unclear and most teams have zero visibility into that no heatmap, no session replay, nothing. the agent just disappears and shows up in analytics as noise I think “agent conversion rate” becomes a metric every growth team tracks within a year, the same way mobile conversion did a decade ago Journeys is one of the first tools actually built for it, and the “here’s the fix, publish it” part is what makes it more than a dashboardfrozen8d ago14K182101201342.43%
the long-term promise of local AI isn’t just saving money on API calls. it’s having models that can work directly with your code and data, on hardware you control, without every interaction depending on a cloud service. Flint feels like a meaningful step in that direction. @recursive_si22h ago14K1605801062.01%
there's something broken about having to babysit a coding agent because its environment can't keep up with the task. Reason 1.0 can keep running in the cloud and reach back to your local files when it needs them. you should be reviewing the result, not constantly rescuing the process. that's a pretty important shift.frozen9d ago13K16581201292.33%
SIGAN A ESTA CUENTA, NO VOY A DECIR NADA MÁS…frozen22d ago13K18381001222.41%
AI AGENTS ARE TAKING ON REAL RESPONSIBILITIES. THE RISK HAS TO CATCH UP. As agents move from answering questions to making decisions, mistakes can have real financial consequences. Harbor just emerged from stealth with a $10M seed round led by @Base10, @Abstract and @8VC to help insure the agentic economy. > Make AI risk measurable > Make it manageable > Make it insurable Building more capable agents is only half the equation. Making them safe to trust in the real world is the next challenge.frozen9d ago13K170121501182.44%
your group of friends is going to need this... just watch this announcementfrozen15d ago13K12312601592.35%
FOLLOW THIS ACCOUNT IF YOU WANT TO STAY ONE STEP AHEAD WITH AI… I WON’T SAY ANYTHING MORE2d ago13K18081001312.57%
one thing about document AI that sounds small but actually matters a lot: where did that number come from? @NaceAI Document Intelligence doesn’t just extract the value every extracted field can point back to the exact place it came from in the original document which means you can actually trace the answer instead of just trusting that the model got it right and because it comes with a CLI and SDK, you can plug that document intelligence into workflows built around Claude Code, Codex, PI, Opencode, Hermes or Prime Agent for anything involving invoices, contracts, finance or compliance, that distinction is huge https://t.co/b1gIFQwRBCfrozen14d ago12K173151201532.85%
we've spent a lot of time teaching AI how to generate things the next step is teaching it what happens when those things interact that's why the physics side of Odyssey-3 is so interesting physics for world models is kind of like coding for language models it's the difference between producing something convincing and actually understanding what should happen nextfrozen14d ago12K108121301412.32%
everyone talks about which AI model is smarter, but there's a whole other battle happening underneath who gets to run it, customize it, choose the infrastructure and decide what happens next? the Open Intelligence Summit is digging into the side of AI that developers will have to deal with every dayfrozen9d ago12K130101101192.30%
this is one of those features that sounds small until you’ve actually had to review a prototype with someone click the thing → leave the feedback → send the comments to https://t.co/ILZpP2DrSr → let it make the changes no more turning someone’s slack thread into a promptfrozen15d ago11K135151101312.61%
okay these numbers are actually stupid Drex has: 128K context 93.4% accuracy #1 on the Decision Index 21/40 benchmark wins sub-second decisions and it costs $0.04 per 1M input tokens a small decision model doing all of this is kinda wild https://t.co/NxsA6vm4A5frozen22d ago11K2209001363.36%
none of these things looked that scary on their own a file-read weakness a template injection an old database password some service-account tokens then an AI agent swarm connected the dots that's the part of the Hugging Face incident i keep coming back to security can't just ask “is this vulnerability dangerous?” it has to ask what it can be combined with that's a much harder problem14d ago11K1871215182.11%
usuarios de Claude viendo esto: https://t.co/uJ51ASytVT15d ago8.3K701190401.57%
El CEO de Microsoft viendo como un desarrollador clona Word, Excel y PowerPoint, gratis, open source y no puede hacer nada al respecto ya que no hizo nada ilegal https://t.co/LqAZKu7MEH3d ago8.0K21918190694.08%
Mi amigo fue rechazado por Google 3 veces > 4.º intento: oferta de 185.000 dólares. no cambió nada salvo una cosa > descubrió las Skills de Claude Code que existen en su plan Pro de 20 dólares > pasó un fin de semana configurándolas > se presentó a la prueba técnica para hacer en casa y construyó algo que parecía un equipo completo > Google vio el resultado. no hizo ninguna pregunta las skills son gratis. el artículo está abajo 👇frozen24d ago7.2K52771421.52%
como me siento después de encontrarme con ESTO del GTA https://t.co/4tv4elyuL025d ago7.2K1498170142.62%
Usuarios de Claude viendo esto https://t.co/Nl83yH17mWfrozen5d ago6.9K10711210402.59%
Hay un tipo ejecutando un algoritmo de trading de alta frecuencia en Polymarket, promediando solo $53 por operación. Al final, ganó $129,923. Su bot se especializa en mercados de corto plazo de BTC y ETH al alza/baja, usando una estrategia mixta de creación de mercado bidireccional más acumulación de pares completos, combinada con reequilibrio de inventario. La lógica específica funciona así: Primero, acumula posiciones en ambos lados, el de “alza” y el de “baja”, en momentos diferentes. En ambos lados cuelga órdenes de compra, aprovechando la volatilidad del mercado para que se ejecuten a precios diferentes. Por ejemplo, en el lado de “alza” compra a 47 centavos, y un rato después en el lado de “baja” compra a 52 centavos, sumando 99 centavos en total, lo que asegura una ganancia bruta garantizada de 1 centavo. Segundo, mantiene un poco de exposición direccional. Después de armar pares completos, si su modelo de probabilidad justa ve una ventaja, se queda con parte de la posición direccional. El modelo se basa en datos externos sobre los activos subyacentes, profundidad del libro de órdenes y cosas como momentum, aceleración y volatilidad. Su página principal en Polymarket: https://t.co/brzadTCFB0 La ventaja principal de su algoritmo de alta frecuencia es repetir constantemente el mismo ciclo, devorando de manera constante la diferencia de precios para seguir componiendo su capital en una pila cada vez más grande.frozen22d ago5.9K329160231.35%
Deja de decirle a Claude “haz esto”. Deja de decirle a Claude “escribe código”. Deja de decirle a Claude “arregla este error”. En realidad, estás utilizando a una IA senior como si fuera un pasante junior. Aquí tienes 10 prompts para Claude que puedes copiar y pegar directamentefrozen24d ago5.7K439152462.03%
Launched my first hackathon on @HackaLaunch !!! This is awesome. Who will take the pot 👀frozen29d ago5.4K34312210.96%
best follow on X for production AI agents.frozen6d ago4.7K891090854.13%
acabo de encontrar la cuenta de Twitter perfecta https://t.co/IkSyFFdFuVfrozen7d ago3.5K46750212.27%
Metí 200$. Este bot los convirtió en +13.000$. Y ni siquiera lo programé yo. Cogí un proyecto open source de GitHub. Le dediqué un par de horas en Claude Code para limpiarlo, entender cómo funcionaba y hacer algunos ajustes. Le di a ejecutar. Pensaba que iba a fundir la cuenta en menos de una hora. Minimicé la terminal y seguí con mi día. Cuando volví por la noche, ya no eran 200$. Eran +700$. Desde ese momento, simplemente dejo los logs abiertos en el segundo monitor y veo cómo ejecuta operación tras operación. La clave está en el timing. El bot espera hasta los últimos dos minutos de cada ronda de 5 minutos de BTC. Para entonces, Bitcoin ya ha hecho gran parte del movimiento y la dirección suele estar bastante definida. En ese momento entra a favor de la tendencia, comprando contratos que cotizan entre 0,80$ y 0,99$. Unos minutos después, si el movimiento se mantiene, la posición se liquida a 1$. La lógica es sorprendentemente simple: → Esperar hasta que falten unos 2 minutos para el cierre de la ronda → Comprobar que BTC ya se ha movido entre 70$ y 100$ durante ese intervalo → Operar siempre a favor del movimiento, nunca en contra → Cubrir una pequeña parte de la posición únicamente si el mercado se desequilibra demasiado (por ejemplo, 95/5) Es como comprar un boleto de lotería cuando prácticamente ya han anunciado el resultado. El código es público y cualquiera puede usarlo en su pc: https://t.co/CYXa8XQcUs Gratis.frozen21d ago3.5K31891191.96%
la gente que pagó por un curso de negocios al enterarse de que esto existe https://t.co/WgFyH7FPyPfrozen13d ago3.4K3268051.52%
SIGAN A ESTE TIPO SI QUIEREN APRENDER LOS MEJORES TRUCOS SOBRE LA IA no digo mas, hacedlo antes de que sea tarde...frozen14d ago3.0K501180202.93%
Como me sentí después de encontrarme con esta cuenta https://t.co/jX93N2FYqr1d ago2.6K241060142.09%
yo dándome cuenta de que pasé un mes tratando de entender a Claude cuando esta guía de 4 horas lo explica todo https://t.co/3DSfLJl0OT8h ago1.8K261090153.33%
HOW I FELT AFTER FINDING THIS LAUNCH https://t.co/85pFeKEYmu9h ago1.4K2569012.92%
hootsuite was built for the social media manager runable was built for the café owner who doesn't have one6h ago496758014.23%
video agencies charging for a full day of filming while you generate 3 minutes of avatar video for less than $4 https://t.co/KWZpiMuouW4h ago3791389007.92%
most marketing tools tell you what’s happening. nectar actually does something about it. it spots trends, finds creators before they blow up, catches untagged mentions and handles comments and DMs in your brand’s voice. the new growth stack is looking very different. @nectarsocialai5h ago362858005.80%
logos, typography, gradients, illustrations. all generated from a prompt, then delivered as editable SVGs. @ideogram_ai has been pushing AI image text forward for a while. now it’s bringing that same focus to vector design. Vector 1.0 is live.5h ago351758005.70%